Decision Making
Probabilistic Selection Task
Also called: PST, choose-A avoid-B task, probabilistic learning task
Measure learning from positive and from negative feedback: participants learn three pairs of symbols from feedback that is right only part of the time, then choose between new pairs, scored as choose-A and avoid-B accuracy.
8-12 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Preview any test free after you sign up.
- Papers, 2021 to 2025
- 31
- Evidence grade
- C
- Registered clinical studies
- 11
What it measures
Measures learning from positive and from negative feedback. Participants learn three pairs of symbols from feedback that is right only part of the time, then choose between new pairings with no feedback. The main score is how often the most rewarded symbol is chosen in the new pairings (choose-A accuracy), with avoidance of the least rewarded one (avoid-B accuracy) next.
Constructs
- Reinforcement learning
- Learning from positive feedback
- Learning from negative feedback
- Approach and avoidance learning
Research uses
- Research on dopamine and reinforcement learning
- Parkinson's disease and psychiatric research
- Within-study comparison of learning from positive and negative feedback
How the paradigm works
The probabilistic selection task measures how much a person learns from positive and from negative feedback (Frank, Seeberger and O'Reilly, 2004). In training, three pairs of symbols, AB, CD and EF, come in random order, and feedback says Correct or Incorrect only part of the time: choosing A is called correct on 80% of AB trials, C on 70% of CD trials and E on 60% of EF trials. A test then pairs the symbols in new ways, with no feedback. Choosing A over other symbols depends on having learned from positive feedback, and avoiding B on having learned from negative feedback. Frank and colleagues found that people with Parkinson's disease learned better from negative feedback off their dopamine medication and better from positive feedback on it.
More about the paradigm
Each trial shows two symbols side by side until the participant chooses one by clicking or tapping, or with the F and J keys, with no time limit; after 4 seconds a line asks for an answer. In training, Correct or Incorrect shows for 1 second with the symbols still on screen. A training block holds each pair 20 times, 10 with each symbol on the left, and the feedback follows fixed counts in each block, so every block has exactly the 80%, 70% and 60% rates. Training ends after the first block from the second on in which A was chosen on at least 65% of AB trials, C on 60% of CD trials and E on 50% of EF trials, or after 3 blocks (2 to 6 in the settings). The test shows all 15 pairs of the six symbols 6 times each, 3 with each symbol on the left, 90 trials in all. Practice comes first: 4 trials with two practice symbols, one always correct. The symbols are irregular shapes drawn by code for this task in place of the source's Japanese characters, all with the same area and ink, and which symbol plays which role comes from the session's seed. Each Correct in training earns 1 point (0 to 10 in the settings; 0 runs without points), and the study decides what points are worth. About 8 to 12 minutes.
The main score is choose-A accuracy: the share of test trials pairing A with C, D, E or F on which A was chosen. Avoid-B accuracy, the share of test trials pairing B with C, D, E or F on which B was not chosen, comes next, then their difference, overall test accuracy, A chosen over B, whether training met the criteria, the blocks it took, the last block's accuracy on each pair, and win-stay and lose-shift rates in training. A session whose training never met the criteria, or in which A was not chosen over B on more than half the AB test trials, is flagged for review. Scores support research and do not diagnose.
Scores
- Choose-A Accuracy%
- Share of test trials pairing A, the symbol called correct most often in training, with C, D, E or F on which A was chosen.
- Higher means more learning from positive feedback; 0.5 is chance
- Avoid-B Accuracy%
- Share of test trials pairing B, the symbol called correct least often, with C, D, E or F on which B was not chosen.
- Higher means more learning from negative feedback; 0.5 is chance
- Choose-A Minus Avoid-B%
- Choose-A accuracy minus avoid-B accuracy.
- Positive values mean more learning from positive than from negative feedback
- Test Accuracy%
- Share of all test choices of the symbol that was called correct more often in training.
- Higher is better; 0.5 is chance
- A Chosen Over B%
- Share of the AB test trials on which A was chosen.
- Half or less is flagged: the pair with the clearest feedback was not learned
- Training Criteria Met0/1
- 1 when training ended on a block, from the second on, in which A was chosen on at least 65% of AB trials, C on 60% of CD trials and E on 50% of EF trials.
- 0 is flagged for review
- Training Blockscount
- Training blocks of 60 trials run.
- Fewer means the criteria were met sooner
- Last Training Block, AB%
- Share of AB trials in the last training block on which A was chosen.
- Higher is better
- Last Training Block, CD%
- Share of CD trials in the last training block on which C was chosen.
- Higher is better
- Last Training Block, EF%
- Share of EF trials in the last training block on which E was chosen.
- Higher is better
- Win-Stay Rate%
- In training, after a choice called correct, the share of next showings of the same pair with the same choice.
- Informational
- Lose-Shift Rate%
- In training, after a choice called incorrect, the share of next showings of the same pair with the other choice.
- Informational
- Points Wonpoints
- Points for training choices called correct.
- Informational
- Median Test Choice Timems
- Median time from the symbols' onset to the choice in the test.
- Informational; the task has no time limit
- Choices After the Remindercount
- Choices made after the 4 second reminder, in training and in the test.
- Informational
- Anticipationscount
- Presses within 100 ms of the symbols' onset. They are not taken as choices.
- Quality check
- Test Choices Madecount
- Test choices in the scored run.
- Fewer than planned is flagged for review
- Planned Test Trialscount
- Test trials in the session: each of the 15 pairs 6 times.
- Informational
Use and evidence
31 papers from 2021 to 2025 name this task in their title or abstract, up 3% compared with 2016 to 2020. That ranks it 298 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 6 |
| 2024 | 8 |
| 2023 | 6 |
| 2022 | 5 |
| 2021 | 6 |
| 2020 | 9 |
| 2019 | 3 |
| 2018 | 7 |
| 2017 | 1 |
| 2016 | 10 |
| 2015 | 2 |
| 2014 | 3 |
| 2013 | 1 |
| 2012 | 6 |
| 2011 | 1 |
| 2010 | 3 |
| 2009 | 1 |
| 2008 | 2 |
| 2007 | 2 |
| 2006 | 0 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 0 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade C
At least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies.
- Papers on test-retest reliability
- 1
- Papers on reliability or validity
- 2
- Registered clinical studies using it as an outcome
- 11
How the grades work
Use: A is 1,000 or more papers from 2021 to 2025, B 300 to 999, C 100 to 299, D 30 to 99 and E fewer than 30. Evidence: A means at least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome; B, at least 8 test-retest papers, and either 5 registered clinical studies or 150 reliability or validity papers; C, at least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies; D, little published measurement evidence found. The counts measure use. They include every paper that names the task in its title or abstract. Grades describe the published research on this task, not ConductCognition's version of it.
Source: ConductCognition test census 0.14.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Symbols
- 6 irregular shapes drawn by code for this task, with the same area and ink, and 2 more for the practice. Which shape plays which role (A to F) is drawn from the session's seed.
- Training pairs
- AB, CD and EF. Choosing A is called correct on 80% of AB trials, C on 70% of CD trials and E on 60% of EF trials; on the other trials the other symbol is.
- Training blocks
- 60 trials: each pair 20 times, 10 with each symbol on the left, the calls fixed in number and spread evenly over the sides, in a seeded order.
- Criteria
- Training ends after the first block from the second on in which A was chosen on at least 65% of AB trials, C on 60% of CD trials and E on 50% of EF trials, or after 3 blocks (standard; 2 to 6).
- Trial
- The symbols stay until a choice by click, tap, or the F and J keys, with no time limit; after 4,000 ms a line asks for an answer. In training, Correct or Incorrect shows for 1,000 ms with the symbols still shown, then a 500 ms blank.
- Test
- All 15 pairs of the six symbols, 6 times each and 3 with each symbol on the left, 90 trials in a seeded order, with no feedback.
- Points
- 1 point (standard; 0 to 10) for each training choice called correct, with the total shown. 0 runs the task with Correct and Incorrect only. The points are hypothetical by default; a setting changes the instruction line to say the study pays.
- Practice
- 4 trials with two practice symbols, one always correct.
Frank and colleagues used Japanese characters, which some participants can read; this test uses shapes of its own, so its accuracies are not directly comparable with theirs. The feedback follows fixed counts within each block rather than a draw on each trial, so every session has the same rates, and the trials have no time limit, where the source studies set one. The scores support research and do not diagnose.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: probabilistic_selection_test from the scored run, a choice between two symbols with no time limit and no feedback. A choice is right when it picks the symbol that was called correct more often in training (chance 50%). Accuracy below chance invalidates and accuracy not above chance is flagged. One side on 90% of choices invalidates when accuracy is not above chance; the better symbol is on each side equally often. Answers under 200 ms on at least half the choices invalidate. Training that ends without meeting the criteria, A chosen over B on half the AB test pairs or fewer, and a run short of its planned choices are flagged for review and withhold the percentile.
Why these rules
Accuracy significantly below chance (one-sided exact test, p < .05) cannot come from guessing; it means the response mapping was reversed or the task was not followed (Slick, Sherman and Iverson 1999 treat below-chance forced-choice performance as definite response bias). Chance-level accuracy can be a real result in impaired participants, so it is only flagged. One option on at least 90% of answered trials, when the correct options are balanced and accuracy is not above chance, means the answers did not depend on the items (the task counterpart of long-string responding in surveys; Curran 2016). An answer under 200 ms is faster than a simple visual reaction, so it cannot reflect the item. Following response-time effort screening for untimed tests (Wise and Kong 2005), the session is invalid only when at least half the answers are that fast. Studies of this task set the learning criteria before the test, and leave out participants who do not choose A over B on more than half the AB test pairs (Schutte and colleagues, 2017), because then the training did not take.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Most training blocks | 3 | 2 to 6 |
| Points per correct | 1 | 0 to 10 |
| Payment line | Hypothetical points (standard) | Hypothetical points (standard), The study pays for points |
Yearly plans can change settings within these limits. Free and pay-as-you-go studies use the standard settings.
Need a setting this test does not have?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Languages: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese.
- Available on: Every plan, including the free plan.
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Probabilistic Selection Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Frank, Seeberger & O'Reilly, 2004), with the standard settings (most training blocks: 3; points per correct: 1; payment line: Hypothetical points). The main score was choose-A accuracy (share of test trials pairing A, the symbol called correct most often in training, with C, D, E or F on which A was chosen).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Frank, M. J., Seeberger, L. C., & O'Reilly, R. C. (2004). By carrot or by stick: Cognitive reinforcement learning in parkinsonism. Science, 306(5703), 1940-1943. https://doi.org/10.1126/science.1102941The task: three pairs at 80/20, 70/30 and 60/40, and the test of new pairs without feedback.
- Frank, M. J., Moustafa, A. A., Haughey, H. M., Curran, T., & Hutchison, K. E. (2007). Genetic triple dissociation reveals multiple roles for dopamine in reinforcement learning. Proceedings of the National Academy of Sciences, 104(41), 16311-16316. https://doi.org/10.1073/pnas.0706111104Dopamine genes and learning on this task.
- Gold, B. P., Frank, M. J., Bogert, B., & Brattico, E. (2013). Pleasurable music affects reinforcement learning according to the listener. Frontiers in Psychology, 4, 541. https://doi.org/10.3389/fpsyg.2013.00541A test of all 15 pairs, 6 times each with 3 in each order, without feedback.
- Schutte, I., Slagter, H. A., Collins, A. G. E., Frank, M. J., & Kenemans, J. L. (2017). Stimulus discriminability may bias value-based probabilistic learning. PLoS ONE, 12(5), e0176205. https://doi.org/10.1371/journal.pone.0176205Blocks of 60 trials, the criteria of 65%, 60% and 50%, and why the symbols' roles should be counterbalanced.
Source papers
- Frank MJ, Seeberger LC, O'Reilly RC (2004). By carrot or by stick: Cognitive reinforcement learning in parkinsonism. Science, 306(5703):1940-1943.N 49, Older adults
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Probabilistic Selection Task take?
About 8 to 12 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Choose-A Accuracy, Avoid-B Accuracy, Choose-A Minus Avoid-B, Test Accuracy, A Chosen Over B, Training Criteria Met, Training Blocks, Last Training Block, AB, Last Training Block, CD, Last Training Block, EF, Win-Stay Rate, Lose-Shift Rate, Points Won, Median Test Choice Time, Choices After the Reminder, Anticipations, Test Choices Made and Planned Test Trials. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Probabilistic Selection Task used in research?
31 papers from 2021 to 2025 name it in their title or abstract, up 3% compared with 2016 to 2020. That ranks it 298 of 395 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.14.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Probabilistic Selection Task have?
In the published research, 1 paper reports its test-retest reliability and 11 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Which settings can a researcher change?
Most training blocks (2 to 6), points per correct (0 to 10) and payment line (Hypothetical points (standard), The study pays for points). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings.
Which languages can participants use?
Participants can take this test in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. English is on every plan, and the other languages come with yearly plans. Have a native speaker read the machine-translated text before launch.
Is this a clinical test?
No. ConductCognition is for research use only. It runs, scores and exports the test; clinical interpretation is outside the platform.
Every test: the full test list.
Run this test in your next study
Start free with 10 free participants on any test, then use participant credits or a yearly plan.